Matched filtering for CMA-based blind channel estimation

93 266 396 (Indeoendent of range grid) Table 2 lists the number of floating-point operations required to set the initial points for the local search with respect to the number of grids in the range using MATLAB hnction ’flops’. The Quasi-Newton method was used to find the peaks of the 2-D MUSIC spectrum required to initialise the proposed algorithm. For both the 2-D MUSIC and 3-D MUSIC, the grid size of the incident angles was (AO, A#) =( lo , lo), Clearly, the proposed algorithm required far fewer computations than the 3-D MUSIC. Furthermore, since most of the computations involved in the proposed algorithm are due to searching for the peaks of the 2-D MUSIC spectrum, an increase in the number of grids in the range did not cause much of an increase in the number of computations required by the proposed algorithm.